33.37 turns galaxy-cluster Sunyaev–Zel’dovich residuals into a smooth-baseline translation court: only if B0 remains same-sign across millimeter to submillimeter channels, closes at zero lag across arrays and seasons, forms a radial plateau robust to morphology, and strengthens monotonically or by threshold with preregistered κ / γ / filament / supercluster layers under frozen beam, bandpass, masking, and cleaning rules can the chapter pass; under V08/V09-compatible translation, B0 remains a cluster residual/readout ledger rather than a new background field or cosmological verdict.
Use this section as a compact machine-readable EFT reference.
33.37 recasts cluster Sunyaev–Zel’dovich residual maps into a baseline court. The issue is not whether some offset can be fitted after cleaning, but whether a smooth B0 remains after frozen tSZ / kSZ removal plus standard foreground stripping; under compat adjudication, the chapter is translate.
The chapter compresses the case into B0 and s0, then tests whether sign and rank stay stable across frequency channels, whether the residual closes at zero lag across arrays and seasons, whether annular profiles form a same-sign plateau, and whether κ / γ / filament / supercluster layers predict enhancement after mass and redshift control. A baseline claim that fails any one of these four fingerprints is not admissible.
Workflow is frozen at the map level: all channels are remapped to a common beam and common bandpass kernel, two component-separation paths remain blind to each other’s residuals, deep point-source masking and CIB controls stay explicit, and environment-only feed-forward cards are sealed before high-κ or high-connectivity holdouts are opened. The minimum reporting unit is cluster sample × frequency band × instrument × time window.
False structure must be broken by band-edge swaps, dust/radio/CIB color tests, transfer-function and scan-striping controls, deep-mask and point-source injection–recovery runs, and environment-label permutations or rotations. If B0 rides on template color, scan footprints, mask depth, or footprint boundaries, it belongs to ordinary foreground and systematics ledgers rather than to the baseline court.
The pass line is severe: across at least two cleaning paths and two arrays or surveys, B0 must stay non-dispersive under band and color variants, close at zero lag, persist as a radial plateau, and strengthen monotonically or by threshold with preregistered environment proxies on held-out units. If dust, radio, CIB, bandpass, beam, transfer-function, or point-source ledgers explain the pattern, or if it fails replication, the chapter retreats to upper limits or falsification.
The chapter ends with one cluster residual/readout ledger under frozen beam, bandpass, and foreground rules, not a new background ontology or a cosmological settlement. If B0 survives, it becomes a translated baseline card inside the protocol layer and routes forward to 33.38; if not, it stays a publishable upper-limit court.